Human–AI Identity Guessing Chat (Imitation Game)

Real-time, turn-based chat game that pairs you with either another user or an LLM bot. After the conversation ends, both sides submit a guess: Human or AI.

Website Hosted

If this is still up, it's hosted on https://bot-or-human-cse-312.onrender.com/

Features

• Real-time messaging over Socket.io with pairing + disconnect handling
• Turn-based chat flow with timers and per-session limits
• AI opponent powered by Google Gemini (Gemma) via an Express API
• Prompt templating + randomized “personality” presets
• Post-chat identity guess collection

Tech Stack

• Frontend: Vite + React + TypeScript, socket.io-client
• Backend: Node.js + Express + Socket.io
• AI: Google Gemini API (Gemma model)

Local Development

Prerequisites

• Node.js (recommended: 18+)

Setup

All app code lives in root/.

1. Install dependencies

cd root
npm install

2. Configure environment variables

Create root/.env (or copy from root/.env.example) and set:

GEMINI_API_KEY=YOUR_KEY_HERE

Run (2 terminals)

Terminal 1 (Socket.io + Express API on port 3001):

cd root
node server.js

Terminal 2 (Vite dev server on port 5173):

cd root
npm run dev

Open the app at the URL printed by Vite (typically http://localhost:5173).

Production Build

cd root
npm run build

Then run the server in production mode so it serves dist/:

cd root
set NODE_ENV=production
node server.js

The server listens on PORT (defaults to 3001).

How It Works (High Level)

• The frontend initiates a session when you click Start Chat.
• For human vs. human sessions, the Socket.io server matches two waiting clients and emits pairing metadata.
• For human vs. AI sessions, the client generates bot responses by calling POST /api/gemini.
• The AI prompt is assembled from:
  • root/src/assets/prompt.txt (template)
  • root/src/assets/personalities.txt (randomized personality blocks)

Socket Events

Client → Server

• choose role — identifies the client role for masking/partner metadata
• join chat — enter the matchmaking queue
• chat message — send a message (server relays and echoes)
• submit guess — submit identity guess after the chat

Server → Client

• waiting — queued for a match
• paired — matched with a partner (includes who goes first)
• chat message — delivered messages (including echo)
• partner guess — partner’s submitted guess
• partner disconnected — partner dropped mid-chat

Configuration Notes

• Vite proxies /api/* to the backend server during development (see root/vite.config.ts).
• Chat limits/timers are implemented as constants in root/src/App.tsx.